You’re Making This Too Hard. AI Just Needs The Data
Getting AI your data isn’t as hard as you think. You’re just not looking in the right place. [Image created by author with Google Gemini.]
The biggest obstacle to turning my ideas into something I can actually work on is usually whether or not I have a pen handy. I like to brainstorm. I like to write. I like to come up with ideas from different angles and unorthodox perspectives and find a way to thread them together. And like many people, a lot of my best thinking comes when I’m not intending to do so. And that’s a problem I’ve struggled to solve.
Everyone has had a brilliant idea in the shower or on the long, boring train ride home that never came to life. A flash of genius that fades because you couldn’t grab onto it long enough to actually write it down. The details are gone, and you’re on to the next thing.
I kept running into this as I started my entrepreneurial journey. I’d have an idea, or an inkling of one at least, and would have to stop what I was doing to try and write it down. If pen and paper wasn’t available, I’d try and capture my thesis via thumbs and screen. Typically it wasn’t enough. I didn’t remember all the nuance and the concept ended before it had begun.
When I started to use AI as a brainstorming partner, its value was hampered by how much detail and context I could feed into it. Despite what some expect, AI cannot read minds. So it was stuck working with two sets of data: my incomplete thoughts and the ideas of a billion people who weren’t me. Neither worked.
So I approached it like I approach most things, from a people and process perspective. I like to break things down into their component parts. Understand what breaks and where, then figure out the why. Obstacles are fine. They are just problems to solve. What obstacles did I have that needed to be overcome?
The first was how do I efficiently and quickly “empty my brain” so that AI could actually do something useful with it? Writing was slow and smartphone notes were worse. Typing was quicker, but I didn’t walk around with a computer all day. I had all these ideas and they seemed stuck somewhere I couldn’t reach.
That’s when I realized something. I had a solution. It was easy, accessible, and best of all, free. If the issue was that I often didn’t have enough time or the proper tools to capture my ideas, that was what I needed to solve for. And the answer to that problem came to me in the drop-off line at my kids’ elementary school.
From Idea to App in One Day
I was sitting in line, waiting for it to move, making sure I didn’t hit a crossing guard or anger a parent already late for work. As I sat there, I had an idea for an app. I wanted to write it down, but crayons and mad libs aren’t the best tools for the job. I assumed I’d lose it like so many other ideas. But that’s when I realized I actually had something that might work. There was a voice recording app on my phone. Not a special AI version. Just the default standard recorder that came on every phone. So I clicked record.
I talked to myself for about 10 minutes in the car. Sometimes with long pauses as I figured out what I was trying to say. Way too many “umm’s” and “uhh’s”. But I captured the idea. It hadn’t been pushed out by other pressing thoughts and distractions. I got it.
When I got home, I opened the file, clicked on the transcript, and exported it to a simple Google Doc. Then I took that transcript and put it into Claude. We went back and forth, exploring ideas, asking questions, researching the competitive landscape, and challenging assumptions. By the end of the day, I had a working prototype in the hands of a client. All from a rambling voice memo in a school drop-off line. (Click here if you want to see a more in depth look at that process.)
I share this story because it highlights a problem that’s bigger than just me and my ideas. AI has a LOT of data in its training models. Enough that it will always answer your question. But if it doesn’t have the right data for your purposes, you get diminished results. A critical factor that determines the quality of your outputs with AI depends on the data you have it use. And unfortunately, they haven’t figured out how to directly beam your thoughts into the screen. That’s a problem, but one that can be overcome. Because there are way more opportunities for you to capture data now than you realize. And they don’t require special software or knowledge. They simply require you to do what you’ve always done, but now with a new set of tools to convert your process into something it can use.
AI Doesn’t Care How It Gets The Data
AI has no preference if the data you give it came from a notebook or a slideshow. It can be a conversation transcript or scribbles on a whiteboard. There is no judgement or complaint about the method you used to collect it. The AI just wants it, so that it can get to work. That’s never been the case before, so we aren’t really thinking about the other options. It was either impossible or required a lot of tech to make it happen. But AI has data capturing capabilities that can massively simplify and accelerate your ability to give it the information it actually needs.
So below are some of my favorite tools and methods to give AI my data. I don’t worry about how pretty and organized it is. I just care that it gets to where I need it to go. On that computer screen. And it doesn’t require you to really do anything new. It just adds a step to make your life that much easier after.
My AI Data Collection Tools And Sample Use Cases
Standard Phone Voice Recorder
Record yourself thinking out loud, transcribed.Turn on the recorder and just go. This isn’t a speech. Style doesn’t matter. The more you ramble, the more data you have. Then download that transcript and go.
Record a conversation, with permission.Some of the best brainstorms come when you’re doing it with someone else. You bounce ideas back and forth, chasing the best ones and seeing where things go. It’s impossible to keep up when you’re taking notes, so don’t. Just hit record.
Record someone else’s answers to a question. Sometimes you need information from someone else, but they either don’t have the energy, time, or ability to write down the answers when and how you need. But answering some questions out loud? Easy enough. Don’t have someone write their job description. Just ask them how they would explain their job to someone else. Then AI takes it from there.
Phone Camera
Snap a photo of your legal pad notes.It’s not always appropriate or useful to record in real time. So if you’re in a meeting or a place where you can’t, go old school. Jot down your notes. Later on you can just take a picture of the page and ask AI to figure it out.
Take a picture of the messy whiteboard session.Building on that brainstorm you had with a colleague, you need to figure out how to get everything from the whiteboard onto a screen. You don’t need the intern to do it anymore.
Upload the image of something without words.Sometimes the data you need isn’t just words. It’s a room you want to design or a place you want to remember. It’s all data.
Snipping/Screenshot Tool
Screenshot the computer screen.“Hey Claude, this was the response from Replit. I have no idea what it means. Can you help me figure it out and what to do next?” Way easier than you trying to copy and paste. Where’s the button? What’s that instruction mean? Screenshot and go.
Screenshot the app you’re building.When you’re creating something new, errors and bugs will always show up. It can be difficult to explain exactly what looks wrong or how you want it fixed. So don’t use words. Screenshot the app, draw a circle around the issue, and tell AI this is what you need solved.
Document Upload
Give the .txt file, word doc, or PDF you need it to read.It knows how.
PowerPoint slides are decipherable by AI.Drop and go.
Excel files aren’t too hard. AI speaks spreadsheet.Tell it where to look and see what it can do.
Comments and mark-ups
This one is worth calling out on its own because of what AI actually does with it. If I have a word document that I made a bunch of corrections and comments on, I upload the whole thing, comments in the margins and all. I tell it to do two things. First, read the original document to know where I was. Then, read the comments to understand what I’ve changed. Not only does AI get your latest thinking, but it can also learn important things from seeing how the idea has evolved. Your AI will be “smarter” because of it.
URLs
You know that article you just read that nailed exactly what you’ve been thinking about? Don’t copy and paste the whole thing. Just grab the URL and give AI its homework assignment. “Read this page.” It pulls the content, processes it, and now it’s part of the conversation. You can ask it to summarize, compare it to something else, or just use it as background while you work on your own thing.
Error Messages
This one’s more technical, but if you’re building anything with AI, error messages are gold. There’s a lot of data packed into them, and AI loves it because they are standardized and consistent. You may not know what every error code means, but AI does. Copy the block of angry red text, paste it in, and let AI translate.
Start Today
That school drop-off line was almost a year ago. Since then, I’ve used every single method on this list to make my process much more simple and valuable. And the thing that still surprises me is how easy doing this actually is. AI doesn’t need a polished document or prose. It needs your ideas, in whatever form you can get them out of your head and onto a screen. Legal pad scrawls and whiteboard doodles and transcribed rambles all do the job.
The next time you have that shower idea or that flash on the train, don’t worry about finding the right way to capture it. There is no right way. Just grab it. AI will figure out the rest.
